I wanted to know more about the BLEU score, so did this experiment:

Test BLEU score

import nltk
reference = "today I go cycling".split()
references = [reference]

hypothesis_strings = ["today I go cycling", 
                      "today I go biking",
                      "today , I go cycling", 
                      "today I go ride my bike", 
                      "today I go bicycles", 
                      "today I go fietsen", 
                      "today I go refrigerator"]
longest_sentence_length = max([len(s) for s in hypothesis_strings])

for hypothesis in hypothesis_strings: 
    bluescore = nltk.translate.bleu_score.sentence_bleu(references,hypothesis.split())
    print("%s %f" % (hypothesis.ljust(longest_sentence_length), bluescore))